Long Lived Transaction Processing for Business Processes and Scientific Workflows

Abstract

Process-oriented systems such as scientific workflows and business processes can be designed and implemented based on advanced transaction models or other formal models for concurrent distributed systems. Advanced transactional modeling approaches trade the classic benefits of isolation and atomicity properties for long lived transactions, giving it the ability of being applicable in systems that run in heterogeneous, autonomous, and distributed computing environments. At same time, process-oriented modeling based on formalisms such as Petri nets, process algebras, and graph-based process models has improved the correctness of designed processes. IME-USP DATA group has been working in important process-oriented and data-oriented challenges which typically have requirements concerning the quality of service and correctness. Among our current research topics, we can highlight: exception handling for long transactions; automated generation of business step dependency representation; data extraction, transformation, and loading for data warehouse systems; automation of clinical and molecular data processing; and formal modeling of business processes aiming qualitative and quantitative analyses.